jpr@jpradley.jpr.com (Jean-Pierre Radley) (02/16/91)
I have a domain. A site in my domain wants to call itself "apple". This is of course a name which already is in the published maps. How do I disambiguate mail destined for apple.jpr.com as opposed to that for apple.com? Jean-Pierre Radley NYC Public Unix jpr@jpradley.jpr.com CIS: 72160,1341
mcn@mimas.UUCP (Michael C. Neuman) (02/17/91)
jpr@jpradley.jpr.com (Jean-Pierre Radley) writes: > I have a domain. > > A site in my domain wants to call itself "apple". > > This is of course a name which already is in the published maps. > > How do I disambiguate mail destined for apple.jpr.com as opposed to that for > apple.com? > I don't think you should have a problem. Since mail destined for apple.com is being sent to a specific domain. Your domain is jpr.com. The only problem would be if you have a link to apple, and it's uucp site name was also apple. THEN you'd have a problem. Mail routing is pretty well based upon domains. For example, my mail feed is bbx.basis.com... My systems address is mimas!xxx@bbx.basis.com... If my feeder were so inclined, I could join his domain and become mimas.basis.com... Or apple.basis.com... Since you have a registered domain, mail sent to jpr.com will go to the system you registered under via the path you registered under. Your system will have to handle the apple.jpr.com conversian to ...!apple. <<<===========================--------==============================>>> <<< Mike Neuman || Senior Systems Programmer >>> <<< mimas!mcn@bbx.basis.com || Albuquerque Academy >>> <<< mcn@beta.lanl.gov || Computer Science Division >>> <<<===========================--------==============================>>> "It's hard to work in a group when you're omnipotent" - Q ST:TNG "Counsel will refrain from making opposing advocate disappear" - Data
zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) (02/18/91)
>> How do I disambiguate mail destined for apple.jpr.com as opposed to that for >> apple.com? >> > I don't think you should have a problem. Since mail destined for >apple.com is being sent to a specific domain. Your domain is jpr.com. The >only problem would be if you have a link to apple, and it's uucp site >name was also apple. THEN you'd have a problem. You don't have a problem until you try to send mail to a uucp site (eg, apple) that has the same name as a host in your domain (eg, apple). Some mailers subscribe to the not very well though out convention of attempting to tack on the local domain (ie, apple!user becomes apple.jpr.com!user). -- Jon Zeeff (NIC handle JZ) zeeff@b-tech.ann-arbor.mi.us
peter@ficc.ferranti.com (Peter da Silva) (02/19/91)
In article <_L=-F2#@b-tech.uucp> zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) writes: > Some mailers subscribe to the not very well though out convention of > attempting to tack on the local domain (ie, apple!user becomes > apple.jpr.com!user). OK, what's wrong with this convention? -- Peter da Silva. `-_-' peter@ferranti.com +1 713 274 5180. 'U` "Have you hugged your wolf today?"
anselmo-ed@CS.YALE.EDU (Ed Anselmo) (02/19/91)
In my ideal mail world:
``apple!user'' refers to UUCP site ``apple'', because the leading
unqualified hostname and the bang implies UUCP routing. On the other
hand ``apple.dom.ain!user'' would be re-written as
``user@apple.dom.ain''.
``user@apple'' would be cannonicalized to ``user@apple.local.domain''
if such a host exists, and then (maybe) handed off to something that
tries some pseudo-domain(s) like .bitnet or .uucp if there's no host
``apple.local.domain''.
--
Ed Anselmo anselmo-ed@cs.yale.edu {harvard,cmcl2}!yale!anselmo-edzeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) (02/19/91)
>> Some mailers subscribe to the not very well though out convention of >> attempting to tack on the local domain (ie, apple!user becomes >> apple.jpr.com!user). > >OK, what's wrong with this convention? When somebody sends mail to mysite!apple!user (intending this mail to be delivered to apple.uucp) it ends up going to user@apple.mysite.com (where it usually bounces). Either get rid of the whole "default domain" thing on sites doing uucp mail or do something like Ed suggested and keep track of the address format and handle uucp differently. The "let's just hope that a name in the default domain doesn't match a uucp name" thing is not the way to do it. -- Jon Zeeff (NIC handle JZ) zeeff@b-tech.ann-arbor.mi.us
fitz@wang.com (Tom Fitzgerald) (02/20/91)
anselmo-ed@CS.YALE.EDU (Ed Anselmo) writes: > In my ideal mail world: > > ``apple!user'' refers to UUCP site ``apple'', because the leading > unqualified hostname and the bang implies UUCP routing. On the other > hand ``apple.dom.ain!user'' would be re-written as ``user@apple.dom.ain''. > > ``user@apple'' would be cannonicalized to ``user@apple.local.domain'' > if such a host exists, and then (maybe) handed off to something that > tries some pseudo-domain(s) like .bitnet or .uucp if there's no host > ``apple.local.domain''. Hmmm, you'd have to be careful about this. When mail gets gatewayed from UUCP to SMTP, a ! address gets automatically changed into a @ address to obey the protocols. When mail gets gatewayed from SMTP to UUCP, the transformation goes backwards, but it's still automatic. You'd have to complicate the transformation some: SMTP->UUCP user@site -> site.local.domain!user rather than site!user UUCP->SMTP site!user -> user@site.UUCP rather than user@site This would certainly have some advantages. The behavior where site!user and user@site are treated identically (which seems to be what most mailers have implemented) causes some problems when a UUCP site name is the same as a machine name in the local domain. --- Tom Fitzgerald Wang Labs fitz@wang.com 1-508-967-5278 Lowell MA, USA ...!uunet!wang!fitz
peter@ficc.ferranti.com (Peter da Silva) (02/23/91)
> >> Some mailers subscribe to the not very well though out convention of > >> attempting to tack on the local domain (ie, apple!user becomes > >> apple.jpr.com!user). > >OK, what's wrong with this convention? > When somebody sends mail to mysite!apple!user (intending this mail to > be delivered to apple.uucp) it ends up going to user@apple.mysite.com > (where it usually bounces). Apple.uucp? Apple.uucp? Where is apple.uucp? I know of an apple.com... is that what you're talking about? If you're planning on letting another site reroute for you, the least you can do is give it an unambiguous address, and "apple" or "apple.uucp" isn't one. Either "apple.com" or "neighbor.dom.ain!apple" where you know that neighbor.dom.ain is adjacent to apple. How do you know that "apple.mysite.com" isn't actually connected to "mysite.com" via a uucp link? In that case the only reasonable interpretation of "apple!anything" is apple.mysite.com. > The "let's just hope that a name in the > default domain doesn't match a uucp name" thing is not the way to do it. I agree. I'd put it differently... the "let's assume that an unqualified ambiguous name is going to be routed correctly" thing is just asking for trouble. If you want to get to a site in the UUCP zone, you have to use a path. -- Peter da Silva. `-_-' peter@ferranti.com +1 713 274 5180. 'U` "Have you hugged your wolf today?"
rickert@mp.cs.niu.edu (Neil Rickert) (02/23/91)
In article <WB?--?&@b-tech.uucp> zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) writes: >When somebody sends mail to mysite!apple!user (intending this mail to >be delivered to apple.uucp) it ends up going to user@apple.mysite.com >(where it usually bounces). If you send 'apple!user' to 'apple.uucp' then you should send 'mysite!apple!user' there also. Otherwise your software is badly broken. If some other site sends mail to you addressed to either 'apple!user' or to 'mysite!apple!user', but you can't get it to the correct destination, then the sending site is badly broken. -- =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= Neil W. Rickert, Computer Science <rickert@cs.niu.edu> Northern Illinois Univ. DeKalb, IL 60115 +1-815-753-6940
scott@skypod.uucp (Scott Campbell) (02/25/91)
In article <1991Feb23.154024.29925@mp.cs.niu.edu> rickert@mp.cs.niu.edu (Neil Rickert) writes: > If you send 'apple!user' to 'apple.uucp' then you should send >'mysite!apple!user' there also. Otherwise your software is badly broken. > > If some other site sends mail to you addressed to either 'apple!user' or >to 'mysite!apple!user', but you can't get it to the correct destination, then >the sending site is badly broken. If you send to apple!user it should go to whoever is listed as apple.uucp If you send to mysite!apple!user it should go to machine apple which talks to mysite (it need not be registered with anyone other than mysite). By explicitly putting mysite in front of apple you specify that mysite should resolve what apple is. This is how you get to unregistered sites. -- Scott J.M. Campbell scott@skypod.uucp Skypod Communications Inc. ..!uunet!scocan!skypod!scott 57 Charles St. West, #1310 ..!uunet!utai!lsuc!becker!skypod!scott Toronto, Ont. (416) 961-3847 ..!epas.utoronto.ca!nyama!skypod!scott
zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) (02/26/91)
>> When somebody sends mail to mysite!apple!user (intending this mail to >> be delivered to apple.uucp) it ends up going to user@apple.mysite.com >> (where it usually bounces). > >Apple.uucp? Apple.uucp? Where is apple.uucp? I know of an apple.com... is >that what you're talking about? Try the uucp maps. #N apple #S DEC VAX-8650; 4.3 BSD UNIX #O Apple Computer, Inc. #C Erik E. Fair #E postmaster@apple.com #T +1 408 974 1779, +1 408 974 3525 #P 20525 Mariani Boulevard, M/S 32-E, Cupertino, CA 95014 #L 122 3 W / 37 22 N It's just an example, replace it with uunet if you prefer. >If you're planning on letting another site reroute for you, the least you >can do is give it an unambiguous address, and "apple" That's the problem. > or "apple.uucp" >isn't one. Wrong. >How do you know that "apple.mysite.com" isn't actually connected to >"mysite.com" via a uucp link? In that case the only reasonable interpretation >of "apple!anything" is apple.mysite.com. Why do you think that it is reasonable to take a well defined address and change it? >> The "let's just hope that a name in the >> default domain doesn't match a uucp name" thing is not the way to do it. > >I agree. I'd put it differently... the "let's assume that an unqualified >ambiguous name is going to be routed correctly" thing is just asking for >trouble. If you want to get to a site in the UUCP zone, you have to use >a path. But you can use a path and still have some sites tack on a local domain (and then send it to the wrong place). -- Jon Zeeff (NIC handle JZ) zeeff@b-tech.ann-arbor.mi.us
rcd@ico.isc.com (Dick Dunn) (02/26/91)
peter@ficc.ferranti.com (Peter da Silva) writes: [background: discussion about tacking domain onto bang-path element] > > >> ...ie[sic], apple!user becomes apple.jpr.com!user... ... > > When somebody sends mail to mysite!apple!user (intending this mail to > > be delivered to apple.uucp) it ends up going to user@apple.mysite.com ... > Apple.uucp? Apple.uucp? Where is apple.uucp? I know of an apple.com... is > that what you're talking about? The maps say that the machine named "apple" in the uucp zone is the same as apple.com. Why is this confusing? > If you're planning on letting another site reroute for you, the least you > can do is give it an unambiguous address, and "apple" or "apple.uucp" > isn't one... How can apple.uucp be ambiguous? That means "machine apple in the uucp zone" and there's only one of those, by definition. >...Either "apple.com" or "neighbor.dom.ain!apple" where you know > that neighbor.dom.ain is adjacent to apple. apple may be a bad example here because it's also in a real domain. What about all the machines which ARE registered in the uucp maps but ARE NOT in a true domain? The sensible approach is either to use the pathalias- derived data, or hand your mail to a machine which does. > ...If you want to get to a site in the UUCP zone, you have to use > a path. It's one thing to insist on a complete path if you're using bang-style source routing; it's another thing entirely to expect the user to create that path. apple!user is a uucp-style path (or fragment of a path). It seems at least reasonable to route as follows: - If you talk to "apple", hand them the mail. - If you don't talk to apple, look up apple in the paths file and route accordingly. - If you don't have the paths on your machine, hand apple!user to a machine that does, and let it route. One way that people get into trouble is with the assumption that host!user is the same as user@host. user@host is a valid way to abbreviate user@host.local.domain. host!user, in a substantial part of the universe, means "`user' on the registered uucp machine `host'". If you DON'T use that convention/mnemonic, how do you (the user) get routing via the maps? You already objected to adding ".uucp" on the end of the hostname. -- Dick Dunn rcd@ico.isc.com -or- ico!rcd Boulder, CO (303)449-2870 ...But is it art?
zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) (02/26/91)
>> >> When somebody sends mail to mysite!apple!user (intending this mail to >> >> be delivered to apple.uucp) it ends up going to user@apple.mysite.com >> >> (where it usually bounces). >> > > >What is "well defined" about the "apple" in "mysite!apple!user", when there >is a private, unadvertised "apple.mysite.com"? Given the remotely >unknowable existence of apple.mysite.com, how can the address be well defined? Let's take a case where mysite advertises (via the uucp maps) a uucp connection to apple.uucp. mysite!apple!user is then quite well defined and yet some mailers will still send it to apple.mysite.com!user. >If you are at mysite.com, and for sufficient reasons "apple.mysite.com" >exists, and the link from "mysite" to "apple.mysite" is via UUCP, and you >receive something for "mysite!apple!user", what should you do? Does mysite show a link to apple.uucp in the uucp maps? Do you want users to be able to rely on pathalias routing to reach apple.uucp? >unique. It would be unreasonable to prohibit local nicnames, as in >user@apple meaning user@apple.mysite.com. I disagree. I do that here and it doesn't cause any problems. >It would be unreasonable to say that "apple!user" is not a legal way for >users at "mysite" to address mail to others at apple.mysite.com, even if >the link is known to be UUCP, and/or NFSNET fair use restrictions require >it not go via the Internet. If mysite has a uucp link to apple.uucp, then I disagree. If not, then it's a local decision. I don't find it unreasonable to make users use apple.mysite.com when that is what they mean. >I think "mysite!apple!user" must mean "deliver to 'user' on the nearby >machine named 'apple'." If the source had meant to send it to apple.com, >then they would have written mysite!apple.com!user or mysite!apple.uucp!user >..!sgi!adobe!apple!user. Certainly if some other site is routing mail to apple.uucp through you (via one hop), they should know what you define "apple" as. You had better define it as "apple.uucp" if you show a link in the uucp maps. Some mailers don't. But what about local users - do you want your users to somehow know that apple is a special case? Or should users always use ".uucp" when they mean a site in the uucp maps? I would find this obnoxious or confusing (foobar!user works but apple!user doesn't because there happens to be a site in the local domain with the name apple but not one named foobar?). -- Jon Zeeff (NIC handle JZ) zeeff@b-tech.ann-arbor.mi.us
vjs@rhyolite.wpd.sgi.com (Vernon Schryver) (02/27/91)
In article <1991Feb26.014736.6765@ico.isc.com>, rcd@ico.isc.com (Dick Dunn) writes: > .... > One way that people get into trouble is with the assumption that host!user > is the same as user@host. user@host is a valid way to abbreviate > user@host.local.domain. host!user, in a substantial part of the universe, > means "`user' on the registered uucp machine `host'". > ... The universe so large in these later days that even a small fraction of the universe is "substantial". However, much of the universe does not use smail, and much of the part that does use smail does not do agressive rerouting. In other words, in an another substantial part of the universe, "user@host" and "host!user" are identical. Many sendmail.cf's immediately convert one to the other. It's hard to have a rational idea of how "most" of the universe treats host!user. My unsupportable intuition is that at most sites "host!user" means what others have said: (1) if "host" = "me", deliver to local "user" or bounce (2) if "host" is a locally know machine, whether on ether, FDDI, UUCP, or jungle drums, send it there. (3) if "host" is in my copy of the maps, and I think I'm a smart host, and it looks like a genuine UUCP path, send it there (the last bit is significant) Vernon Schryver, vjs@sgi.com
vjs@rhyolite.wpd.sgi.com (Vernon Schryver) (02/27/91)
In article <8YF-!-?@b-tech.uucp>, zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) writes: > ...[things that I mostly agree with].... > > > It would be unreasonable to prohibit local nicnames, as in > >user@apple meaning user@apple.mysite.com. > > I disagree. I do that here and it doesn't cause any problems. What!? You are able to make people type "host.ann-arbor.mi.us"? That boggles my mind. It seems most of the thousands of people at the thousands of machines within sgi.com complain bitterly at having to type "host.subdom" instead of "host". They'd get out their pitchforks and torches if told to type "host.subdom.ain.sgi.com"! Could it be there are few machines in ann-arbor.mi.us, so that people rarely have to type the extra stuff? Here, everyone has at least one private machine, and most of us spend much of every day fiddling with many other machines, frequently many routers and domains distant. > If mysite has a uucp link to apple.uucp, then I disagree. If not, then it's > a local decision. I don't find it unreasonable to make users use > apple.mysite.com when that is what they mean. It seems strange to advertise a link to "apple" meaning apple.com without having such a link. It seems crazy to both advertise such a wierd link and route mail elsewhere, so you and I must agree. > But what about local users - do you want your users to somehow know > that apple is a special case? Or should users always use ".uucp" when > they mean a site in the uucp maps? I would find this obnoxious or > confusing (foobar!user works but apple!user doesn't because there happens to > be a site in the local domain with the name apple but not one named foobar?). Here, sgi!foo... becomes foo... immediately foo!user is immediately converted to user@foo foo.domain!user is immediately converted to user@foo.domain foo!bar!user is treated like foo!bar.uucp!user (see @bar.uucp below) foo!bar.uucp!user is treated as if it were an RFC822 source route through bar.uucp (see @bar.uucp below) user@bar is routed to: (1) bar.sgi.com if existent (2) directly connected UUCP neighbor if existent else (3) bounced user@bar.uucp is routed where the maps say bar exists, regardless of whether there is also a local UUCP neighbor foo or a foo.sgi.com. user@bar.uux is routed to the directly connected UUCP neighbor bar or bounced. This is for UUCP neighbors who can't pick unique good names. Notice that we don't do UUCP map lookups on unqualified hostnames unless there is a strong hint that UUCP is involved. By going to FQDM's early and well away from the UUCP & Internet gateways, people rarely notice & never complain of the obnoxiousness you mention. Moreover, people here have come to rarely use simple UUCP hostname paths. Instead, they use domains, as in user@apple.com, even for UUCP links. Vernon Schryver, vjs@sgi.com
zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) (02/27/91)
>of machines within sgi.com complain bitterly at having to type >"host.subdom" instead of "host". They'd get out their pitchforks and >torches if told to type "host.subdom.ain.sgi.com"! Yes, it's an inconvenience. There are aliases for frequent recipients and most mail is via "reply". Otherwise, maybe a centralized mail server (so all sgi users just type "user") is the way to go. It makes it much less confusing - I hate questions like "why do I send to bill with just bill, chris with chris@site, and jan with jan@site.domain". "Why does jan tell me that everyone else can reach her with "jan@site" and yet I can't". I also hate to constantly have to remember which machine I'm on, ie do I need to use joe, joe@site, or joe@site.domain? Depends on which window I'm in. >Here, > foo!user is immediately converted to user@foo > user@bar is routed to: > (1) bar.sgi.com if existent > (2) directly connected UUCP neighbor if existent > else (3) bounced I assume that you also implement the policy that you won't have any external uucp neighbors that have a uucp name the same as one of your local names? Otherwise, someone offsite sending mail (routed via the uucp maps) might send to sgi!foo!user and not have it end up where they expected. That was the original point, and unless I missed something, you could still suffer from this problem. -- Jon Zeeff (NIC handle JZ) zeeff@b-tech.ann-arbor.mi.us
peter@ficc.ferranti.com (Peter da Silva) (02/28/91)
In article <8YF-!-?@b-tech.uucp> zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) writes: > Let's take a case where mysite advertises (via the uucp maps) a uucp > connection to apple.uucp. Let's not. Seriously, if they've got a direct external link through UUCP to apple they should not have an internal site with the same name. If someone does that, then got on their case. > mysite!apple!user is then quite well defined > and yet some mailers will still send it to apple.mysite.com!user. Mailers where? At mysite? If mysite has a local system named "apple" that is the system they *should* send it to. If you're outside mysite and you expect it to route properly to apple!user, when they haven't advertised such a link, then you deserve to lose. > >If you are at mysite.com, and for sufficient reasons "apple.mysite.com" > >exists, and the link from "mysite" to "apple.mysite" is via UUCP, and you > >receive something for "mysite!apple!user", what should you do? > Does mysite show a link to apple.uucp in the uucp maps? No. And if it's an internal site it doesn't need to. > Do you want > users to be able to rely on pathalias routing to reach apple.uucp? No, I want them to be able to rely on pathalias routing to reach apple.com. > If mysite has a uucp link to apple.uucp, then I disagree. If not, then it's > a local decision. I don't find it unreasonable to make users use > apple.mysite.com when that is what they mean. No, local users should be able to assume that ambiguities will be resolved in favor of the local machine. That's the common case. > But what about local users - do you want your users to somehow know > that apple is a special case? Or should users always use ".uucp" when > they mean a site in the uucp maps? No, I don't tell my users about the UUCP maps. I tell them to use fully domainised names for sites outside Ferranti. Then if it breaks I fix it. If they want to bang-route stuff themselves that's their responsibility. -- Peter da Silva. `-_-' peter@ferranti.com +1 713 274 5180. 'U` "Have you hugged your wolf today?"
vjs@rhyolite.wpd.sgi.com (Vernon Schryver) (02/28/91)
In article <NNG-J*B@b-tech.uucp>, zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) writes: > Yes, [typing FQDM's] an inconvenience. There are aliases for frequent > recipients and most mail is via "reply". Otherwise, maybe a centralized > mail server (so all sgi users just type "user") is the way to go. Or a common, global, distribute name space. (Not to open the perennial centralized/distributed "discussion") > >[rules at at sgi.com include:] > > foo!user is immediately converted to user@foo > > user@bar is routed to: > > (1) bar.sgi.com if existent > > (2) directly connected UUCP neighbor if existent > > else (3) bounced > I assume that you also implement the policy that you won't have any > external uucp neighbors that have a uucp name the same as one of your local > names? Otherwise, someone offsite sending mail (routed via the > uucp maps) might send to sgi!foo!user and not have it end up where > they expected. That was the original point, and unless I missed > something, you could still suffer from this problem. Yes: I think it would be silly, ridiculous, wrong, impolite, and a sign of Bad Breeding to advertise in the UUCP maps a route to one place and send mail somewhere else. No: mail to sgi!<commonname>!user, where <commonname> is not in the sgi map entry and is in the sgi.com domain will go to the sgi host. Also No: if one of our map neighbors decides to add an alias in their UUCP map entry that happens to be the same as an internal machine, then mail sent to that alias will lose. Vernon Schryver, vjs@sgi.com
rpw3@rigden.wpd.sgi.com (Rob Warnock) (03/02/91)
In article <88013@sgi.sgi.com> vjs@rhyolite.wpd.sgi.com (Vernon Schryver) writes: +--------------- | > >[rules at at sgi.com include:] | > > foo!user is immediately converted to user@foo | > > user@bar is routed to: | > > (1) bar.sgi.com if existent | > > (2) directly connected UUCP neighbor if existent | > > else (3) bounced | > I assume that you also implement the policy that you won't have any | > external uucp neighbors that have a uucp name the same as one of your local | > names? Otherwise, someone offsite sending mail (routed via the | > uucp maps) might send to sgi!foo!user and not have it end up where | > they expected... | | Yes: I think it would be silly, ridiculous, wrong, impolite, and a | sign of Bad Breeding to advertise in the UUCP maps a route to one | place and send mail somewhere else. | | No: mail to sgi!<commonname>!user, where <commonname> is not in the sgi | map entry and is in the sgi.com domain will go to the sgi host. +--------------- Actually, it's a little better than Vernon implies. The machine "sgi" (a.k.a. sgi.com, sgi.sgi.com, & sgi.uucp) is a firewall; it actually knows very little about SGI machines other than specific internal mail relays. Thus while a conflict *is* resolved the way Vernon says, the *chance* of a conflict is minimized, since most local machines are not in fact directly visible to "sgi". For example, my machine at home is named "redwood" (a.k.a. "redwood.uucp", registered in the UUCP maps long before I came to work for SGI). There is also an internal machine "redwood" (a.k.a. redwood.wpd.sgi.com). Anywhere on the internal net, mail to "user@redwood" will go to the internal machine. On "sgi", mail to "user@redwood" will go to my home machine (as it should, since "sgi" advertises that in the UUCP maps), and mail to "user@redwood.wpd.sgi.com" will go to the internal machine. It works. -Rob ----- Rob Warnock, MS-1L/515 rpw3@sgi.com rpw3@pei.com Silicon Graphics, Inc. (415)335-1673 Protocol Engines, Inc. 2011 N. Shoreline Blvd. Mountain View, CA 94039-7311
peter@ficc.ferranti.com (Peter da Silva) (03/05/91)
In article <J_F-9G-@b-tech.uucp> zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) writes: > Try the uucp maps. I'm going to have to replace my supply of smileys. Yes, I know "apple" is in the uucp maps. But the site name is "apple.com", and depending on "apple" or any other name in the maps to be unique... except among other systems in the uucp maps... is just asking for trouble. Yes, using "something.uucp" to refer to the maps is common, but is it actually officially supported? Who is the contact for the ".uucp" domain? > >How do you know that "apple.mysite.com" isn't actually connected to > >"mysite.com" via a uucp link? In that case the only reasonable interpretation > >of "apple!anything" is apple.mysite.com. > Why do you think that it is reasonable to take a well defined address and > change it? But "apple!something" is not a well-defined address. If I'm at foo.bar.com and I have a uucp link to the sales office in Apple City, and their uucp name is "apple", then what should I do with "apple!something"? > But you can use a path and still have some sites tack on a local > domain (and then send it to the wrong place). Could you give an example of this behaviour? -- Peter da Silva. `-_-' peter@ferranti.com +1 713 274 5180. 'U` "Have you hugged your wolf today?"
peter@ficc.ferranti.com (Peter da Silva) (03/05/91)
In article <1991Feb26.014736.6765@ico.isc.com> rcd@ico.isc.com (Dick Dunn) writes: > >...Either "apple.com" or "neighbor.dom.ain!apple" where you know > > that neighbor.dom.ain is adjacent to apple. > apple may be a bad example here because it's also in a real domain. What > about all the machines which ARE registered in the uucp maps but ARE NOT > in a true domain? The sensible approach is either to use the pathalias- > derived data, or hand your mail to a machine which does. OK, now you're assuming that "mysite" does this. I'm just pointing out that you can't assume that J. Random System out there in the net will be able to accept a path starting with an unqualified host name and route it appropriately. Now if someone at *mysite* is routing like this and gets a bounce from apple.mysite, they can presumably get better routing info from their system admins. But third parties can't assume that. > One way that people get into trouble is with the assumption that host!user > is the same as user@host. This assumption is built into too much software to change. > You already objected to adding ".uucp" on the end of the hostname. user@apple.uucp is as good a way of doing it as anything. The basic problem is that there really is no good way. How do folks out on the internet get mail to uucp sites? Most of the ones I know just punt it to a site that they know does pathalias. How about formalising this behaviour and making uucp a real domain with all these known smart sites as MX? -- Peter da Silva. `-_-' peter@ferranti.com +1 713 274 5180. 'U` "Have you hugged your wolf today?"